EC2 Right-Sizing Estimator
Enter your current instance type and actual CPU/memory utilisation to get your optimal instance recommendation — and see exactly what you'd save.
Your Current Setup
Select your instance type and enter your average utilisation over the past 30 days.
Current Instance Specs
Right-Sizing Analysis
| Instance | vCPU | RAM | On-Demand ($/hr) | Monthly Cost |
|---|
Save More With Scheduled Downtime
Right-sizing is a one-time win. Scheduling your instances off during off-hours cuts costs by 60–70% every month — automatically.
Try ServerScheduler Free →Most AWS accounts have at least a handful of EC2 instances that are larger than they need to be. The estimator above helps you model what moving those instances to a smaller type would cost, based on current on-demand pricing. Use it alongside CloudWatch CPU and memory metrics to build a prioritised list of right-sizing candidates.
What Is EC2 Right-Sizing
Right-sizing means choosing an EC2 instance type whose CPU, memory, network, and storage specifications match what your workload actually uses rather than what you conservatively provisioned. Overprovisioning is extremely common — instances are frequently launched with the next size up "just in case," and then the workload never grows into that capacity.
AWS charges the same hourly rate whether an instance is at 5% CPU utilisation or 95%. A well-provisioned instance running at 70% utilisation costs the same as an oversized one running at 10%. The difference is that the second scenario is wasting money that could be recovered simply by choosing a smaller type.
AWS and third-party cost management reports consistently find that 30 to 50 percent of EC2 instances in typical accounts are oversized by at least one size. In accounts with a long history and no systematic right-sizing programme, that figure is often higher.
Identifying Oversized Instances
The standard method for identifying right-sizing candidates is to look at average and maximum CPU utilisation over a rolling 14-day or 30-day period in CloudWatch. Instances with average CPU below 20% and maximum CPU below 50% are strong candidates — there's clearly headroom to move down a size without risking performance.
Memory utilisation is harder to assess because CloudWatch doesn't collect it by default. You need the CloudWatch agent installed and a custom metric configured. For instances where you can't measure memory, a conservative approach is to only right-size when CPU utilisation strongly supports it, and monitor the instance after the change.
| Avg CPU | Max CPU | Right-Sizing Signal | Recommended Action |
|---|---|---|---|
| < 10% | < 30% | Strong candidate | Downsize one size immediately |
| 10–20% | 30–50% | Likely candidate | Downsize one size with monitoring |
| 20–40% | 50–70% | Review memory | Downsize only if memory allows |
| > 40% | > 70% | Appropriately sized | No action needed |
Choosing the Right Instance Family
Right-sizing isn't only about moving to a smaller instance within the same family. Switching instance families can unlock significantly better price-performance. The m5 family is a good general-purpose baseline, but compute-optimised (c-family) instances offer better CPU per dollar for CPU-bound workloads, and memory-optimised (r-family) instances offer better memory per dollar for memory-heavy applications.
Newer generation instances — m6i, m7i, c6i — are also typically 5 to 10% cheaper than their previous-generation equivalents at the same size. If your account has older generation instances, migrating to the current generation is a straightforward saving even without changing the size. ServerScheduler's EC2 right-sizing recommendations surface both size and generation opportunities automatically.
Risks to Manage When Downsizing
The main risk with right-sizing is degraded performance if the new instance type turns out to be undersized for peak loads. This is manageable with a few precautions. Schedule the resize during a low-traffic window — nights or weekends for non-production, a maintenance window for production. Monitor the instance for 24 to 48 hours after the change, with alerting on CPU and memory. And have a documented rollback path: resizing back up takes just a few minutes if you need to.
ServerScheduler's scheduled resize feature lets you automate this: define the new instance type, pick a maintenance window, and the platform handles the stop, resize, and restart. If something goes wrong, you can define an automatic revert. This turns a manually risky operation into a controlled, repeatable process.
If you're unsure whether a right-sized instance will perform adequately, test it in your dev or staging environment first. Apply the same schedule to a copy of the workload and run your test suite against it before touching production.
Pairing Right-Sizing With Scheduling
Right-sizing and instance scheduling are complementary strategies. Right-sizing reduces the cost per hour of each instance. Scheduling reduces the number of hours each instance runs. Applied together, the savings multiply. A right-sized dev environment running on a standard 8am–8pm weekday schedule typically costs 75 to 80% less than an oversized environment running around the clock.
Use the AWS Cost Savings Calculator to model the scheduling component, and this estimator to model the right-sizing component. Combine both figures to understand the full optimisation opportunity in your account.
Get automated right-sizing recommendations from ServerScheduler.
Scheduled EC2 ResizeAutomate instance resizes during maintenance windows.
Cost Savings CalculatorModel savings from scheduling alongside right-sizing.
How to Reduce AWS CostsA full guide to cutting costs across your AWS account.